The graphics and statistical information on this page fill gradually as they become available, with some not available until the next day. The page is updated every 30 minutes at about 20 and 50 minutes past the hour.

For weather news as it breaks that is tagged and organised, use the links on the Weather and Climate Media Reports page.

Monday 17 April 2017

 Deadly dilemmas: why does the forecaster get blamed when the forecast goes wrong?

It's likely there will be criticism of New Zealand's MetService because it over-forecast the strength of ex-cyclone Cook when it hit the country last Thursday and Friday. Certainly, the dire predictions being made by Metservice and the Civil Defence, and published by a voracious media, would not have lined up with what most Kiwis saw outside their windows. Yes, there was certainly transport disruption as well as some damage and flooding in the east and some central parts of the North Island, but for much of the country that had been led to expect the biggest event since the Wahini Storm of 1968, "Cyclone Cook" was pretty halfhearted.

MetService forecaster John Crouch told Radio New Zealand the cyclone had been difficult to predict because it was such an intense system, and computer modelling does not always get that exactly right.

"And because it's been quite small and compact, [and changed its] track by about 50 to 100km it's made quite a difference to the impact on the country. What we have found is that the system has tracked a little further east than we initially thought, and the central pressure hasn't been quite as deep as we were initially thinking."

Those two factors resulted in the impact of the ex-cyclone on the two largest centres in the country, Auckland and Wellington, being well below expectations. Aucklanders, who had been exhorted to stay indoors and cancel their Easter holiday activities would not have been sympathetic as they contemplated the sunny, perfect holiday weather on Saturday. Of course, they might have been thankful that the worst case scenario - being flooded and blown to smithereens - had not eventuated, but there's no escaping the fact that the wrong forecast was on very public display to well over 50% of the NZ population.

Many of these good people have little interest in the weather, except when it affects them or their plans, and won't have either listened to or understood MetService's explanation. They just put it down to "MetService got it wrong...again" and move on. This could be regarded as just a piquing of MetService's pride except that there is something much more important at stake.

When ex-Hurricane Sandy blasted the east coast of the United States late in October 2012 causing the deaths of 157 people and $US71.4 billion in damage (second only to Hurricane Katrina), a remarkable 54% of vulnerable residents in Sandy's path did not comply with weather warnings. Research after the event showed that this was due in part to the "cry wolf" effect of past experience of false alarms. Perceived wrong forecasts don't just have an effect at the time, but create a legacy of mistrust that can be deadly in future events.

"Blame the meteorologists!" as Bryan Norcross exclaims in this blog on "Rethinking Big Storm Warnings and Guaranteed Handwringing" following the fizzer-forecast for city-stopping blizzards in New York last month. Immediately after the (non) event, New Jersey Governor Chris Christie was angry, and that was his answer: Blame the meteorologists. Schools, some roads and public transport had been closed down and a smalll fortune spent on salt in anticipation of the event. But slightly warmer air moved into the area pushing the rain/snow line a little west so that, while areas and major cities farther inland got the forecast blizzards and heavy snow, New York only experienced rain or sleet.

Unfortunately, unless perfect science can be conjured up, there will continue to be uncertainties and therefore forecast errors. Some things are just beyond our knowledge and ability to forecast, especially small-scale events and the precise boundaries of larger damaging systems. For example, we can say with some certainty whether an area will have thunderstorms, and even whether they will be severe, but knowing where an individual storm will strike, much less if a tornado will develop and where, are beyond us. We are getting much better at predicting the behaviour of cyclones, though the smaller they are the harder the call. Rainfall amounts, and where the heaviest rain will occur, are still challenging to forecast as is the location of the rain/snow line when the surface temperature will be close to freezing.

Bryan Norcross again: "You would think that someone who has lived in New Jersey as long as the governor would understand the rain-snow line dilemma. If an unforecastable difference in the storm track means somewhere between debilitating and nuisance, you’d better err on the side of shutting things down. Unless, of course, you don’t mind the New Jersey Turnpike looking like the Long Island Expressway did in 2013. Thousands of stranded people in stuck cars and trucks, and a cleanup nightmare that took days to resolve."

The forecaster will continue to get the full blame for actions taken in response to a warning unless the responsibility is shifted to those who take actions on the basis of the warning. And for that to be meaningful, the warning must contain information on uncertainty. Respected severe weather meteorologist, Chuck Doswell, puts it well in this article on "What does the public want from a weather forecast?"

"Every forecast that doesn’t include uncertainty information is tantamount to withholding critical information from the public! And the public needs to accept some responsibility to learn how to use that uncertainty for their own purposes – they have to set their own thresholds regarding uncertainty. If the worst thing that could happen to you is getting a little wet, you can accept more uncertainty than if you stand to lose your life if some hazardous weather potential exists."

And what of the fulminating Governor Christie, with a major city to manage? Norcross's solution is "a map with contours that show where there’s a 10%, 20%, etc. chance of the Blizzard Warning being required. 

"Get buy-in in advance from everybody involved, including the loudmouth politicians, on what the odds should be when the trigger is pulled, and the NWS is largely out of the blame game. If the governor is happy to shut down the turnpike if there is at least a one-in-five chance of thousands of his residents getting stranded, set the threshold at 20%. It’s a simple risk-tolerance exercise [that] puts decision-makers’ skin in the game."

In Australia, we have adopted probability forecasts for rainfall in daily forecasts: "there's an 80% chance of rain, most likely 10mm but there's a 25% chance of getting 20mm"). Some warnings are written in a conditional way: "Severe thunderstorms are likely to produce damaging winds, and heavy rainfall that may lead to flash flooding. Storms could cause DAMAGE TO HOMES AND PROPERTY", but here uncertainty is to cover the impossibility of forecasting the locations individual storms will strike - there's no indication of how certain the forecaster is that the overall severe weather event will occur. Many others, such as cyclone warnings, use an unconditional wording with the fear being that complexity will cause confusion.

And that's the dilemma. Do you make a black-and-white statement that some event will happen, and get the entire blame if the warning is wrong? Or do you deduce the most likely scenario from a complex picture using all available information and present probabilities to a diverse public so they are better informed of the realities?

It's a dilemma that some national meteorological agencies are working on but none seem yet to have found the magic bullet. The US National Weather Service is trialling maps contoured with percentage probabilities of fire dangers and snowfall depths being exceeded, while closer to home, MetService produces severe weather risk outlook maps of New Zealand showing its confidence levels that rain, wind or snow events will occur.

Apart from injecting reality into warnings, a move to risk-based wording would reduce the unwarranted hard time forecasters are subjected to. Norcross: "Now we hear, 'The National Weather Service says we’re getting two feet of snow so we’re shutting down the city.' The new paradigm would result in something like, 'The storm threat has reached the level where we can’t take a chance on people being stranded and infrastructure being damaged, so were closing down.'"

 Darwin shivers in exceptional weather set-up

The denizens of Darwin felt unaccustomed cold last Tuesday 21 April as an unusual set of meteorological conditions kept temperatures low. The highest the mercury reached in the 24 hours to 09.00 on 22 April was 21.9°, a massive 2.7° below the previous coldest April maximum temperature in 71 years of record. It was 10.8 below normal and also the coldest maximum for any Wet Season and the third coldest on record for any month. The temperature dropped to 18.0° at 20.15 on 21 April, the coldest April minimum since 1948.

BoM senior forecaster Craig Earl-Spurr told ABC News the cold weather was due to a low pressure system sitting off the NT coast. "The south-easterlies on the southern side of the low were still bringing in quite a bit of dry air near the surface but we had rain coming in from higher up, which kind of worked like an evaporative air-conditioner where the rain coming into the drier air cools that air," Mr Earl-Spurr said. "It is not something you see very often, so it is pretty remarkable."

Many other centres in the Top End had a record cold day in one way or another. Pirlangimpi, right beside the sea on the Tiwi Islands, 130km NNW of Darwin, recorded a top of 23.6°, nearly two degrees below its previous coldest April day and 0.6 below its previous all-time record in 35 years of record-keeping. Middle Point, Batchelor, Dum in Mirrie and Douglas River all with shorter climate histories, also set new April records while Point Fawcett on the Tiwi Islands not only set a new all-time record but also knocked 4.0° off its previous coldest April day in 18 years.

Recent weather briefs - AFRICA

• Severe flooding in Zimbabwe between January and April, including from Tropical Cyclone Dimeo in March, has affected 36 of the country's districts and led to 251 deaths and the destruction of nearly 2,600 homes, according to the country's UN Resident Co-ordinator.. 388 schools have been destroyed and thousands are displaced, with typhoid and malaria an increasing problem acounting for 194 deaths. 140 dams have been breached and there is major damage to infrastructure making many areas inaccessible. Combined with the effects of recent and ongoing drought, over 5.2 million people are affected. The government had declared a national emergency and is appealing for $US189 million in international aid.
• Floodlist reports that flooding in Malawi, NE of Zimbabwe, in early April has left four dead, three missing as at 6 April, 5,520 households affected, 30,000 evacuated and 1,075 hectares (10.75sq km) of crops damaged.
• The drought in the Horn of Africa, covering Ethiopia, Somalia, Kenya and Uganda, threatens the lives of 16 million people reports Al Jazeera if rains fail in 2017 as they did in the 2016 March to June long rainy season and the short wet season in October and November. The Horn suffers the flip side of the Indian Ocean Dipole, which gave Australia rain during 2016 and early this year from warm waters in the northern Indian Ocean, but brought cooler than normal waters and dry weather to NE Aftica. Climate change has brought a steady reduction in rain during the long rainy season but a majority of climate models are predicting a return to average or above average falls for the long rains this year